Research map / Probabilistic and scientific modeling / Molecules and proteins
GLIME: General, Stable and Local LIME Explanation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use GLIME-Binomial instead of ALIME | ImageNet explanations with VGG16, σ=0.25 and 128 samples · ImageNet | Top-20 Jaccard index: 0.618 -> 0.952 | Helped |
| Use GLIME-Gauss instead of ALIME | ImageNet explanations with VGG16, σ=0.5 and 128 samples · ImageNet | Top-20 Jaccard index: 0.525 -> 0.875 | Helped |
| Use GLIME-Gauss instead of GLIME-Binomial | ImageNet explanations with VGG16, σ=0.5 and 128 samples · ImageNet | Top-20 Jaccard index: 0.596 -> 0.875 | Helped |
| Use GLIME-Gauss instead of ALIME | ImageNet explanations with VGG16, σ=1 and 128 samples · ImageNet | Top-20 Jaccard index: 0.519 -> 0.883 | Helped |
| Use GLIME-Binomial instead of ALIME | ImageNet explanations with VGG16, σ=5 and 128 samples · ImageNet | Top-20 Jaccard index: 0.489 -> 0.493 | No clear effect |
| Use a local, reference-independent sampling distribution instead of LIME sampling | GLIME explanations · ImageNet | local fidelity | Helped |
| Integrate LIME's weighting function into the sampling distribution | LIME via GLIME-Binomial · ImageNet | stability: 2000+ samples -> 200-400 samples | Helped |
| Remove LIME's weighting function (ablation) | LIME · ImageNet, especially small σ | stability | Helped |
| Remove LIME's regularization (ablation) | LIME · ImageNet, especially small σ | stability | Helped |
| Vary the sampling distribution among Gaussian, Laplace, and Uniform (ablation) | GLIME · ImageNet local neighborhoods | local fidelity | Mixed |
| Use GLIME instead of LIME | Explanations of accurate predictions · Human interpretability ratings | participant rating: 2.96 -> 3.37 | Helped |
| Use GLIME instead of LIME | Explanations of incorrect predictions · Human interpretability ratings | participant rating: 2.33 -> 3.42 | Helped |
Closest papers by meaning
- SLICE: Stabilized LIME for Consistent Explanations for Image Classification (CVPR 2024 · similarity 0.87)
- Regional Explanations: Bridging Local and Global Variable Importance (NeurIPS 2025 · similarity 0.85)
- Additive Models Explained: A Computational Complexity Approach (NeurIPS 2025 · similarity 0.83)
- Using Stratified Sampling to Improve LIME Image Explanations (AAAI 2024 · similarity 0.83)
- Trust Regions for Explanations via Black-Box Probabilistic Certification (ICML 2024 · similarity 0.83)
- Efficient and Accurate Explanation Estimation with Distribution Compression (ICLR 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.